Re: ipkdb

Kamil Rytarowski <[email protected]> Mon, 9 Oct 2017 13:46:10 +0200

Newsgroups	gmane.os.netbsd.ports.i386
Message-ID	<[email protected]>

On 09.10.2017 13:29, Maxime Villard wrote:
> Hi,
> is there any interest in maintaining ipkdb on i386? It is poorly
> implemented:
>  * Until I removed vm86, ipkdb couldn't interact well with userland 8086
>    applications.
>  * It does not support SMP.
>  * What ipkdbsetup and friends do is basically crazy (patching the IDT
> etc).
> 
> I'm trying to simplify the entry/exit points in the kernel, and the
> presence of
> ipkdb does not help. I would like to remove it from i386. Note also that
> there
> is no support for it on amd64.
> 
> Maxime

Hello,

Yes, there is interest to resurrect ipkdb(4).

Revisit of ipkdb(4) has been roughly scheduled after finishing LLDB,
ptrace, sanitizers, (+kernel-sanitizers?), restoration of GDB and
aligning kgdb to the reality... before that it's not the proper time to
touch it, at least by myself.

I plan to research current usability and how to make this kernel feature
more MI, at least compatible with any network device.

src/blob/trunk/doc/TODO.ptrace: - research ipkdb(4)

When I raised questions about the current ipkgdb(4) implementation
Christos suggested to leave it as it is to keep as a reference for
future work.

signature.asc (application/pgp-signature, 819 B)